Geoscience Reference
In-Depth Information
FIGURE 10.3
Example of a stratified sample of 36 sites from three strata. In each stratum, 12 sites were
selected using the simple random design.
When constructing strata, it is best to base strata boundaries on factors that
either do not change or change slowly. Examples of good strata boundar-
ies are geographic (mountain ranges, elevation, etc.) or political boundaries
(states, counties, etc.). Examples of poor strata boundaries are those based
on habitat classifications or distances from roads. These criteria make poor
strata boundaries because they have the potential to change rapidly.
10.4.1.4 Cluster Sampling
Cluster samples are special cases of two-stage samples that occur when all
secondary units from selected primary units are selected. Like two-stage
sampling, cluster sampling draws a sample of primary units at stage 1. At
stage 2, however, a cluster design selects all secondary units within each
selected primary. An example cluster sample is shown in Figure 10.4.
In ecological studies, cluster sampling can be useful when all sample units of a
collection (cluster) are relatively easy to collect. For example, in some situations it
may be relatively easy to capture an entire group of gregarious animals (e.g., wild
horses, dolphins, etc.). In these cases, it is reasonable to define the group as a clus-
ter and individuals within the group as the secondary units. Cluster sampling
is not popular for sampling geographic locations, except for the special case of
systematic sampling, unless the number of clusters sampled is high.
